Attribution models: Which marketing channel generates sales?
Potential clients embark on a long journey before deciding to buy a product. They search for information via price comparison websites, go to shops, watch commercials, visit your website and weigh up all the alternatives before finally making a purchase. By leveraging on your (omni-channel) marketing channels, you can influence this journey to (hopefully) ensure that your shop is the final destination. Your main objective is of course to use your limited marketing budget in the most effective way possible. And that raises the question as to which channel will optimise your sales?
Determine the added value of your channels
Marketeers are using attribution models for valuation of their various marketing channels and find out which one contributes most to the eventual conversion/sale. To determine this, general methods such as ‘first click’ or ‘last click’ attribution are used. These methods are also included in Google Analytics. As the name indicates, the methods attribute all the value of a conversion to the first or last interaction with the client. Another attribution method is 'position-based'. But is this what you want? After all, these methods entirely ignore the effects of any interim interactions with the customer.
This issue was previously discussed in the blog by my colleague Wouter Hosman in which he also offers a solution: use a statistical model to gain insight into the effects of your omni-channel marketing activities.
Using a statistical attribution model
While insights provided by standard attribution methods are initially useful, they lack subtlety as they do not take into account all other channels. Or the contribution of a channel is based on arbitrary rules. A statistical model determines a weight or value to all included channels, based on objective data. These can include both online and offline channels. The value indicates the extent to which the channel contributes to conversion. These values can then be directly applied in analytical tools, such as Google Analytics.
In addition, statistical models use all the valuable information hidden in your Big Data. And not only information about individuals who have converted. The model also looks at those that have not (yet) converted. Why didn’t they convert? Which channels did they visit? This is all important information.
Another benefit of statistical models is that you can update it constantly with new available data. This way, the scores are always based on the most recent information, and allow you to apply the results in (near) real-time on an individual customer level. Say someone received an e-mail and saw a banner – is another banner or Facebook Ad the best follow-up communication for them? The model provides statistical rules which allow the marketing team to take immediate action. The algorithms enable the optimisation of the customer journey per client.
Your next step in creating attribution models?
Are you already using attribution models or looking to start? Remember the benefits of statistical models. Let the model use the information concealed in your data to enable you to determine the optimal values of your online and offline channels. This way, you will know the extent to which they contribute to your sales and make the most of your marketing budget!